Good times. Does this mean that there are a lot of HTML parser tests which are also failing? Is it safe to assume that this problem would go away with Gerrit, which is based on Git instead of patch files?

It depends on whether the problem is with the patch storage in Rietveld, or the patch application in the bot_update step. Whichever it is, it's probably also the same problem that's causing  bug 656171 .

Possible next steps to investigate this:
 1. Try to make test CLs on both Rietveld and Gerrit with various line endings, and triggering try jobs. If patch failures happen on both Rietveld and Gerrit, this suggests that the failure might be related to patch application.
 2. If the failure is related to patch storage/application that's not specific to Rietveld, then maybe the next place to look is build/scripts/slave/bot_update.py or depot_tools/apply_issue.py.

Meanwhile, until this is solved, the temporary workaround is to skip some tests :-/

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8b137c363843aa9d1bb7783bb364a0584e238d6f

commit 8b137c363843aa9d1bb7783bb364a0584e238d6f
Author: qyearsley <qyearsley@chromium.org>
Date: Fri Feb 10 23:41:57 2017

W3CImportExpectations: Skip webvtt/webvtt-file-format-parsing/webvtt-file-parsing/001.html.

R=maksim.sisov@intel.com
BUG= 691107 

Review-Url: https://codereview.chromium.org/2690633004
Cr-Commit-Position: refs/heads/master@{#449789}

[modify] https://crrev.com/8b137c363843aa9d1bb7783bb364a0584e238d6f/third_party/WebKit/LayoutTests/W3CImportExpectations

Is this resolved if we have .gitattributes to indicate the file is not text?

I think it would -- good idea :-D

In other particular cases where there needs to be various different line endings in the test file (anything with CR AT EOL in https://github.com/w3c/web-platform-tests/blob/master/lint.whitelist), that may also work.

Do you think it would make more sense to put this .gitattributes file upstream, or in Chromium (e.g. in LayoutTests/external/)?

I thought .gitattributes in the upstream.
However, WPT importer should be robust, and detecting CR line endings and adding .gitattributes automatically would be another option.

Note: if this isn't solved after migrating to Gerrit ( bug 693792 ), adding entries in .gitattributes would be a good next thing to try.

Haven't seen this in a while, not sure if it's still an issue, closing bug for now.